A new kind of autonomous discovery

Biological evolution is an autonomously improving system. It builds tools — from neural circuits to limbs — to produce an increased ability to understand and act in the world.

Cultural evolution is another autonomously improving system. It too builds tools — from the termite-fishing sticks built by chimpanzees to the planes built by humans — however, it has scaled an order of magnitude faster because its tools evolved within a lifetime instead of across generations.

These phenomena express a universal learning dynamic — a recursive computation through which matter organizes itself to better predict and shape its environment.

Today, we believe it is possible to build a new kind of autonomously improving system that scales even faster by increasing the bandwidth between experimentation, learning, and embodiment. Leveraging advances in machine learning and physical tools, we aim to solve the grandest challenges of the universe — across physics, chemistry, and biology.